Enforce Mandatory AI Guardrails Across All Agents
AI Guardrails launched as an opt-in feature, a great starting point for exploration. But as organizations move toward full-scale production, the need for consistent, company-wide standards has grown. We’re now making it easier for administrators to set a universal safety foundation, ensuring that every agent aligns with your compliance goals.
Mandatory AI Guardrails
System Administrators can now define a Mandatory AI Guardrails configuration for each deployment stage: DEV, TEST, and PROD. Once defined, the configuration applies automatically to every agent execution in that stage. Developers cannot disable or bypass it.
Developers retain control at the agent level, but only in one direction. The system enforces a security ratchet: a developer can tighten an enabled category (changing PII Filtering from Log to Mask or Block, for example) but cannot reduce enforcement below the organizational baseline. This lets teams increase safety for sensitive use cases without ever undermining the floor the administrator set. Execution diagnostics reflect the effective merged configuration, making it straightforward to test and troubleshoot guardrail behavior across both layers. AI Guardrails is now generally available and you can configure them by going to your agents directly to set them up.
Find out more about Agent Guardrails- ODC
- Artificial Intelligence
